app.py

import json
import pandas as pd
import requests
from multiprocessing import Pool
from functools import partial
import streamlit as st


GITHUB_CODE = "https://huggingface.co/datasets/lvwerra/github-code"
INCODER_IMG = (
    "https://huggingface.co/datasets/loubnabnl/repo-images/raw/main/incoder.png"
)
MODELS = ["CodeParrot", "InCoder"]

@st.cache()
def load_examples():
    with open("utils/examples.json", "r") as f:
        examples = json.load(f)
    return examples


def generate_code(model_name, gen_prompt, max_new_tokens, temperature, seed):
    url = (
        f"https://hf.space/embed/loubnabnl/{model_name.lower()}-subspace/+/api/predict/"
    )
    r = requests.post(
        url=url, json={"data": [gen_prompt, max_new_tokens, temperature, seed]}
    )
    generated_text = r.json()["data"][0]
    return generated_text


st.set_page_config(page_icon=":laptop:", layout="wide")

# Introduction
st.title("Code generation with 🤗")
with open("utils/intro.txt", "r") as f:
    intro = f.read()
st.markdown(intro)

# Pretraining datasets
st.title("1 - Pretraining datasets 📚")
st.markdown(
    f"Preview of some code files from Github repositories in [Github-code dataset]({GITHUB_CODE}):"
)
df = pd.read_csv("utils/data_preview.csv")
st.dataframe(df)
st.header("Model")
selected_model = st.selectbox(
    "Select a code generation model", MODELS, default=["CodeParrot"]
)
with open(f"datasets/{selected_model.lower()}.txt", "r") as f:
    text = f.read()
st.markdown(text)

# Model architecture
st.title("Model architecture")
st.markdow("Most code generation models use GPT style architectures trained on code. Some use encoder-decoder architectures such as AlphaCode.")
st.header("Model")
selected_model = st.selectbox(
    "Select a code generation model", MODELS, default=["CodeParrot"]
)
with open(f"architectures/{selected_model.lower()}.txt", "r") as f:
    text = f.read()
st.markdown(text)
if model == "InCoder":
    st.image(INCODER_IMG, caption="Figure 1: InCoder training", width=700)

# Model evaluation
st.title("Code models evaluation 📊")
with open("evaluation/intro.txt", "r") as f:
    intro = f.read()
st.markdown(intro)

# Code generation
st.title("Code generation 💻")
st.header("Models")
selected_models = st.sidebar.multiselect(
    "Select code generation models to compare", MODELS, default=["CodeParrot"]
)
st.header("Examples")
examples = load_examples()
example_names = [example["name"] for example in examples]
name2id = dict([(name, i) for i, name in enumerate(example_names)])
selected_example = st.selectbox(
    "Select one of the following examples or implement yours", example_names
)
example_text = examples[name2id[selected_example]]["value"]
default_length = examples[name2id[selected_example]]["length"]
st.header("Generation settings")
temperature = st.slider(
    "Temperature:", value=0.2, min_value=0.0, step=0.1, max_value=2.0
)
max_new_tokens = st.slider(
    "Number of tokens to generate:",
    value=default_length,
    min_value=8,
    step=8,
    max_value=256,
)
seed = st.slider(
    "Random seed:", value=42, min_value=0, step=1, max_value=1000
)
gen_prompt = st.text_area(
    "Generate code with prompt:",
    value=example_text,
    height=220,
).strip()
if st.button("Generate code!"):
    with st.spinner("Generating code..."):
        # Create a multiprocessing Pool
        pool = Pool()
        generate_parallel = partial(
            generate_code,
            gen_prompt=gen_prompt,
            max_new_tokens=max_new_tokens,
            temperature=temperature,
            seed=seed,
        )
        output = pool.map(generate_parallel, selected_models)
        for i in range(len(output)):
            st.markdown(f"**{selected_models[i]}**")
            st.code(output[i])